1 / 37

Reusable Fuzzy Extractors for Low-Entropy Distributions

This work explores key derivation from noisy sources with low entropy, aiming to provide secure and stable output. New constructions and limitations of traditional approaches are discussed.

Download Presentation

Reusable Fuzzy Extractors for Low-Entropy Distributions

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Benjamin Fuller Joint work with Ran Canetti, Omer Paneth, Adam Smith, and Leonid Reyzin Reusable Fuzzy Extractors for Low-Entropy Distributions May 9, 2016

  2. High-entropy sources are often noisy • Initial reading initial≠ later reading final • Source initial= a1,…, ak, each symbol ai over alphabet Z • Assume a bound on Hamming distance: d(initial, final) ≤ t Goal: derive a stable cryptographically strong output • Give same key if d(initial, final)<t • Key appears (pseudo)random to adversary Key Derivation from Noisy Sources Physically Unclonable Functions (PUFs) [PappuRechtTaylorGershenfeld02] Biometric Data Challenge: Traditional approaches do not provide adequate security for many physical sources

  3. Key Derivation from Noisy Sources Modeling Noisy Sources and Defining Security Limitations of Traditional Approaches New Constructions Outline

  4. Parameters: How far apart are initial and final? Noise or t (from experimentation) How many values of initial are there? Define entropy as k=log(# values of initial) Quality of a noisy source initial – create a new person, take initial reading <t final–new reading from same person

  5. Key Derivation from Noisy Sources Interactive Protocols[Wyner75] … [BennettBrassardRobert85,88] …lots of work… Today: focus on noninteractive setting, user authenticating to insecure device initial final Parties agree on cryptographic key

  6. Fuzzy Extractors [JuelsWattenberg99], …, [DodisOstrovskyReyzinSmith04] … key • Enrollment Gen: Create keyfrom initial, also create public value help. • Authentication Rep: produce key if d(initial, final) < t • Assume attacker knows distribution of initial • Security: attacker shouldn’t know anything about key key Rep Gen initial help < t final

  7. Fuzzy Extractors [JuelsWattenberg99], …, [DodisOstrovskyReyzinSmith04] … key • Enrollment Gen: Create keyfrom initial, also create public value help. • Authentication Rep: produce key if d(initial, final) < t • Security: keylooks random key Rep Gen Want security when device state is compromised initial help < t final

  8. Fuzzy Extractors [JuelsWattenberg99], …, [DodisOstrovskyReyzinSmith04] … key • Enrollment Gen: Create keyfrom initial, also create public value help. • Authentication Rep: produce key if d(initial, final) < t • Security: keylooks random even if attacker knows help key Rep Gen Want security when device state is compromised initial help < t final

  9. Fuzzy Extractors [JuelsWattenberg99], …, [DodisOstrovskyReyzinSmith04] … key • Enrollment Gen: Create keyfrom initial, also create public value help. • Authentication Rep: produce key if d(initial, final) < t • Security: keylooks random even if attacker knows help key Rep Gen Security can be information-theoretic or computational [DodisOstrovskyReyzinSmith04], [FullerMengReyzin13] initial help < t final

  10. Key Derivation from Noisy Sources • Model and Defining Security • Limitations of Traditional Approaches • Many biometrics [BlantonHudelson09]and PUFs [KoeberlLiRajanWu14]have little entropy, traditional fuzzy extractors provide no security • New Constructions Outline Is it possible to provide security for low entropy sources? Consider case where entropy k < t (number of errors)

  11. Is it possible to handle low entropy: k < t ? Possible values of initial

  12. Is it possible to handle low entropy: k < t ? Possible values of initial t final Recall: attacker can run Rep on any point

  13. Is it possible to handle low entropy: k < t ? Possible values of initial t final Recall: attacker can run Rep on any point

  14. Is it possible to handle low entropy: k < t ? Possible values of initial t key r Rep Recall: attacker can run Rep on any point

  15. Is it possible to handle low entropy: k < t ? Possible values of initial t key r Rep Key derivation is not possible if the source is distributed like this Recall: attacker can run Rep on any point

  16. Is it possible to handle low entropy: k < t ? Counterexample goes away if source distributed like this! Possible values of initial • Checking in with reality: • Iris has: • |initial|= 2048 bits • k= 249 • t≈300 • Closest distance for two people is 680 bits[Daugman04] t final Noisy sources have more structure than k and t !! Recall: attacker can run Rep on any point

  17. Strong Authentication through Key Derivation • Model and Defining Security • Limitations of Traditional Approaches • New Constructions • Use properties of distribution beyond entropy (k) • Observation: many noisy sources are redundant; subsampling increases entropy rate Outline

  18. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations key initial= a1 a2 a3 a4 a5 a6 a7 a8 a9 Gen help r r r r r r r r r r r r a7 a5 a6 a2 a8 a7 a1 a9 a2 a3 a5a2 a3 a9 a5 a3 a4 a5

  19. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations key initial= a1 a2 a3 a4 a5 a6 a7 a8 a9 Gen help r r r r r r r r r r r r a3 a4 a5 a2 a8 a7 a1 a9 a2 a7 a5 a6 a3 a9 a5 a3 a5a2

  20. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations key initial= a1 a2 a3 a4 a5 a6 a7 a8 a9 Gen help r r r r r r key key key key key key a1 a9 a2 a3 a9 a5 a3 a5a2 a2 a8 a7 a7 a5 a6 a3 a4 a5

  21. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations - = locks + positions of symbols to unlock help key initial= a1 a2 a3 a4 a5 a6 a7 a8 a9 Gen help r r r r r r a2 a8 a7 a3 a5 a2 a7a5a6 a3 a4 a5 a3 a9 a5 a1 a9 a2

  22. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations - = locks + positions of symbols to unlock help key initial= a1 a2 a3 a4 a5 a6 a7 a8 a9 Gen help r r r r r r a2 a8 a7 a3 a5 a2 a7a5a6 a3 a4 a5 a3 a9 a5 a1 a9 a2

  23. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations - = locks + positions of symbols to unlock help help r r r r r r a3 a5 a2 a7a5a6 a3 a4 a5 a3 a9 a5 a2 a8 a7 a1 a9 a2

  24. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations - = locks + positions of symbols to unlock help Rep: final= a1 a2a3 a4a5 a6 a7a8 a9 key Rep help r r r r r r a1 a9 a2 a2 a8 a7 a7a5a6 a3 a4 a5 a3 a9 a5 a3 a5 a2

  25. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations - = locks + positions of symbols to unlock help Rep: final= a1 a2a3 a4a5 a6 a7a8 a9 key Rep help r r r r r r a3 a9 a5 a3 a5 a2 a2 a8 a7 a7a5a6 a3 a4a5 a1 a9 a2

  26. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations - = locks + positions of symbols to unlock help Rep: Use the symbols of finalto open at least one lock final= a1 a2a3 a4a5 a6 a7a8 a9 key Rep help r r r r r r a1 a9 a2 a3 a4a5 a2 a8 a7 a3 a5 a2 a3 a9 a5 a7a5a6

  27. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations - = locks + positions of symbols to unlock help Rep: Use the symbols of finalto open at least one lock final= a1 a2a3 a4a5 a6 a7a8 a9 key Rep help r r r r r r key a2 a8 a7 a3 a9 a5 a1 a9 a2 a3 a4a5 a7a5a6 a3 a5 a2

  28. Idea: “encrypt” key using parts of initial Gen: - get random combinations of symbols in initial key - “lock” r using these combinations - = locks + positions of symbols to unlock help Rep: Use the symbols of finalto open at least one lock Error-tolerance: supports any t=o(|initial|) Security: must be hard to guess each combination r r r r r r key a3 a9 a5 a3 a4a5 a7a5a6 a3 a5 a2 a2 a8 a7 a1 a9 a2

  29. Must be hard to guess each combination • Necessary (+sufficient): Samples from source have super logarithmic entropy • Includes: • All high entropy sources • Low entropy sources where symbols are redundant • Subset of all sources where initial are far apart in Hamming space Security r r r r r r key a3 a9 a5 a1 a9 a2 a3 a4a5 a7a5a6 a2 a8 a7 a3 a5 a2

  30. Must be hard to guess each combination • Necessary (+sufficient): Samples from source have super logarithmic entropy • Includes: • All high entropy sources • Low entropy sources where symbols are redundant • Subset of all sources where initial are far apart in Hamming space Security Need to show how locks are implemented r r r r r r key a3 a9 a5 a1 a9 a2 a3 a4a5 a7a5a6 a2 a8 a7 a3 a5 a2

  31. A lock is the following program: • If input = a1 a9 a2, output key • Else output  • One implementation: lock = key || 00…00 Hash(a1 a9 a2)(secure in the random oracle model) How to implement locks? • Ideally: Obfuscate this program • Obfuscation: preserve functionality, hide the program • Obfuscating this specific program called “digital locker” key a1 a9 a2

  32. Digital Locker is obfuscation of • If input = a1 a9 a2, output key • Else output  • Equivalent to encryption of keythat is secureeven when encrypted multiple times with correlated, weak keys [CanettiKalaiVariaWichs10] • Digital lockers are practical(random oracle, discrete logarithm-based) [CanettiDakdouk08], [BitanskyCanetti10] • Hides keyif input can’t be exhaustively searched(superlogarithmic entropy) Digital Lockers key a1 a9 a2

  33. Q: if you are going to use obfuscation, why bother?Why not just obfuscate the following program for help • If distance between initialand input is less than t, output key • Else output  • A: you can do that [BitanskyCanettiKalaiPaneth14], it is: • inefficient • requires very strong assumption • not known to be reusable Digital Lockers • Digital Locker is obfuscation of • If input = a1 a9 a2, output key • Else output  • Equivalent to encryption of r that is secureeven multiple times with correlated and weak keys[CanettiKalaiVariaWichs10] • Digital lockers are practical (R.O. or DL-based) [CanettiDakdouk08], [BitanskyCanetti10] • Hides r if input can’t be exhaustively searched(superlogarithmic entropy) key a1 a9 a2

  34. Same source can be enrolled multiple times with independent services A Reusable Fuzzy Extractor key initial Gen help Secret even given help, help', help'',key, key'' key' initial' Gen help' key'' initial'' Gen help''

  35. Same source can be enrolled multiple times with independent services In the past: difficult to achieve, because typically new enrollments leak fresh information (esp. in help) Only previous construction [Boyen2004]:each enrollment must differ by fixed constants (unrealistic) Our construction: digital lockers leak no partial information (w.h.p.), only requirement: sampling condition holds for each reading A Reusable Fuzzy Extractor

  36. Questions? r r r r r r key a3 a9 a5 a1 a9 a2 a3 a4a5 a7a5a6 a2 a8 a7 a3 a5 a2

More Related